Support Vector Machine Based Classification of Students’ Mathematics Learning Preferences Using Educational Data Mining

Abstract

Educational institutions increasingly seek data-driven approaches to understand students’ learning preferences, particularly in mathematics education where differences in conceptual understanding across domains may influence learning outcomes. This study aims to develop a machine learning–based model to identify students’ mathematics learning preferences across Algebra, Geometry, and Statistics using questionnaire data. The research was conducted at Universitas PGRI Adi Buana Surabaya involving undergraduate students in mathematics education. A quantitative approach was employed using questionnaires as the primary instrument to capture students’ learning preference characteristics. The dataset was analyzed using the Support Vector Machine algorithm with a hold-out validation technique, where the data were divided into 75% training data and 25% testing data. Model performance was evaluated through multiple experimental trials using accuracy, precision, recall, and F1-score. The results show that the model achieved strong and stable performance across ten trials, obtaining an average accuracy of 97.05%, precision of 94.88%, recall of 98.32%, and F1-score of 96.39%. Domain-level analysis indicates that the model performs most consistently in identifying Algebra and Statistics preferences, while slightly greater variation occurs in the Geometry domain due to overlapping conceptual characteristics. Overall, the model successfully detects patterns in questionnaire responses and classifies students’ learning preferences across the three domains. The novelty of this study lies in integrating questionnaire-based educational preference analysis with a machine learning classification framework to automatically detect dominant mathematics learning domains. The implication of this research highlights the potential use of machine learning to support data-driven decision-making and personalized learning strategies in mathematics education. The contribution of this study is the development of a practical approach for applying computational classification techniques to educational datasets in order to better understand students’ learning characteristics and support adaptive learning environments.